Adding duration-based quality labels to learning events for improved description of students’ online learning behavior

EasyChair Preprint 1037

Abstract

Many existing studies analyzing log data from online learning platforms model events such as accessing a webpage or problem solving as simple binary states. In this study, we combine quality information inferred from the duration of each event with the conventional binary states, distinguishing abnormally brief events from normal or extra-long events. The new event records, obtained from students’ interaction with 10 online learning modules, can be seen as a special form of language, with each “word” describing a student’s state of interaction with one learning module, and each “sentence” capturing the interaction with the entire sequence. We used second order Markov chains to learn the patterns of this new “language,” with each chain using the interaction states on two given modules to indicate the interaction states on the following two modules. By visualizing the Markov chains that lead to interaction states associated with either disengagement or high levels of engagement, we observed that: 1) disengagement occurs more frequently towards the end of the module sequence; 2) interaction states associated with the highest level of learning effort rarely leads to disengaged states; and 3) states containing brief learning events frequently lead to disengaged states. One advantage of the current approach is that it can be applied to log data with relatively small numbers of events, which is common for many online learning systems in college level STEM disciplines. Combining quality information with event logs is a simple attempt at incorporating students’ internal condition into learning analytics.

Keyphrases: Log data analysis, Markov chain, online learning modules

@booklet{EasyChair:1037,
  author    = {Matthew Guthrie and Zhongzhou Chen},
  title     = {Adding duration-based quality labels to learning events for improved description of students’ online learning behavior},
  howpublished = {EasyChair Preprint 1037},
  year      = {EasyChair, 2019}}